1. Field of the Invention
The present invention generally relates to a method and an apparatus for estimating troubles in a network, and more particularly, to the estimating of a trouble portion in the network including network apparatuses and connections.
2. Description of the Related Art
Conventionally, troubles in a network system are resolved in the following manner:
(1) When a trouble occurs, various network apparatuses transmit alarm signals. The alarm signals are collected using a monitoring system, for example.
(2) Generally, plural alarm signals may be collected. Only the alarm signals transmitted by the network apparatus that has actually caused the trouble or is close to the actual cause are extracted, and the other alarm signals transmitted by the other network apparatuses that have been affected by the actual cause are discarded. This process is performed by a main cause determination system or manually by an operator, for example.
(3) The actual cause of the trouble is further pursued and estimated based on the extracted alarm signals with reference to past experiences and on trial and error basis. This process is performed by operators.
(4) The cause of the trouble is removed, and the network system is recovered by replacing the package of a causing network apparatus and/or repairing connections.
The above steps (1) and (2) in which the alarm signals are collected and extracted may be supported by a system. Conventionally, the above step (3) is not supported by a system in most cases, and rather performed by operators based on their experiences.
In some cases, a knowledge management system storing past trouble instances has been used for supporting the estimation of cause by the operators. However, such knowledge management system stores instances using natural sentences, and as a result, the operators need to retrieve the instances by natural sentences.
Japanese Patent Laid-Open Application No. 5-114899 discloses a network trouble remote monitoring system that estimates trouble portion in a network by identifying a predefined set of possible trouble portions for each alarm signal based on the alarm signals, the connective relation of the network apparatuses, and a network connection configuration model, and obtaining a product of the identified predefined sets.
Japanese Patent Laid-Open Application No. 7-245609 discloses a network trouble remote monitoring system that uses a trouble propagation model created based on the design knowledge of network elements constituting the network thereby to analyze the cause of trouble based on symptoms.
Japanese Translation of Published PCT International Application No. WO94/19888 discloses a network trouble remote monitoring system that retrieves a trouble ticket associated with unsolved trouble ticket using an information theory derived algorithm and neural network learning algorithm, for example.
Japanese Patent Laid-Open Application 5-260049 discloses a network trouble remote monitoring system that, when an alarm signal is transmitted, estimates which, malfunction or the influence of the malfunction, the alarm signal is caused by, based on the relative probabilities in which a channel or the apparatus to which the channel is accommodated malfunctions or is influenced by the malfunction for each trouble reason.
If the trouble cause is estimated by operators using their human skills and experiences, since each operator has different skill and experience, not all of them can estimate the trouble cause quickly and correctly.
In case of the conventional knowledge management system storing past trouble instances, most of them registers knowledge using natural sentences, and retrieves the knowledge by natural sentences. As a result, such a knowledge management system fails to uniquely represent the portion causing an alarm signal, symptoms, and trouble causes. Even if a similar trouble occurs at another portion, the knowledge management system cannot retrieve appropriate past trouble instance.